From form to function: an ontology of communicative influence beyond truth and intent
This paper proposes a truth- and intent-agnostic ontology for analyzing communicative influence at scale, which structures messages into four orthogonal dimensions (Method, Narrative, Polarization, and AboutObject) grounded in observable cues to support both close reading and large-scale computational analysis of strategic communication beyond traditional disinformation frameworks.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the vast, noisy landscape of modern communication, a single image or a short video clip can move millions of people to action, often without saying a single lie. Consider a photograph of a crowded border crossing. The image itself is real; the people are there, the fence is there, and the sky is exactly as it appears. Yet, when that photo is shared with a specific caption or placed next to a particular headline, it stops being just a picture of a scene and becomes a tool. It can signal danger, define who belongs and who does not, and urge a specific political response. This is the realm of communicative influence, a field where the power of a message lies not in whether its facts are true or false, but in how it is built to shape what we think, how we feel, and what we do. For decades, researchers have focused heavily on spotting lies and identifying bad actors, but a new wave of influence campaigns uses true facts, emotional framing, and clever storytelling to achieve their goals. As artificial intelligence begins to generate these messages at a speed and volume no human team can read, the old ways of analyzing them are no longer enough. We need a new map to understand the terrain.
A team of researchers has drawn that map. They have created a detailed system for breaking down any piece of communication to see exactly how it works, regardless of whether it is telling the truth or trying to deceive. Instead of asking "Is this fake?" or "Who made this?", their new system asks four specific questions about the message itself: How does it try to influence us? How is its story organized? How does it position different groups of people against or alongside each other? And what broader outcome is it trying to achieve? By answering these questions, the researchers can describe the inner machinery of a message, separating the method of persuasion from the story being told, and the group dynamics from the ultimate goal. This approach treats the message as a self-contained object, analyzing the visible clues within it rather than guessing at the secret intentions of the author or the truth of the claims.
The researchers built this system on two simple but powerful ideas. First, they assume that when someone tries to influence an audience, they leave observable traces in the words, images, and structure of the message. Just as a carpenter leaves sawdust and tool marks, a communicator leaves specific patterns that reveal their strategy. Second, they recognize that influence does not require lying. A message can be entirely factually accurate and still be a powerful tool for manipulation if it selects certain facts to highlight, frames them in a specific way, or activates deep-seated group identities. This means that a system designed to catch liars will miss a vast amount of influence work that operates entirely within the bounds of the truth. The new framework is designed to catch both, focusing on the form of the message rather than its factual accuracy or the hidden motives of its creator.
To make this work on a massive scale, the researchers divided the analysis into four distinct dimensions that operate independently of one another. The first dimension, which they call the method, looks at the specific techniques used to sway an audience. This includes things like using loaded language, simplifying complex causes, or appealing to fear. The second dimension, narrative, examines how the message organizes its story. It looks at who is cast as the hero or the villain, how time is sequenced, and what moral stakes are raised. The third dimension, polarization, focuses on how the message positions groups of people. It identifies whether the message is trying to divide people into "us versus them," bring a group together, or bridge a gap between different sides. The final dimension, the influence object, identifies the broader outcome the message is aiming for, such as changing a policy, boosting voter turnout, or eroding trust in an institution. Crucially, these four dimensions are separate; a message can use the same method to tell a very different story, or it can tell the same story while aiming for a completely different result.
The team applied this system to real-world examples, including a famous political campaign image from the United Kingdom that showed refugees at a border. Even though the image contained no false statements, the system revealed how it worked. The method used was a threat frame, positioning the scene as a danger to the nation. The narrative cast the European Union as an external enemy and the local people as a group under siege. The polarization dimension showed how the message divided the world into a vulnerable "us" and a threatening "them." Finally, the influence object was clearly identified as a push for a specific political vote. By separating these elements, the researchers could see exactly how a true image was being used to drive a specific political outcome, a nuance that would be lost if the analysis only looked for lies or tried to guess the author's intent.
To ensure the system was robust, the researchers checked that each of the four dimensions was truly necessary and could not be replaced by the others. They showed that knowing the method used does not tell you the story being told, and knowing the topic does not tell you the goal. For instance, two messages about immigration could use the exact same emotional tactics, but one might aim to increase voter turnout while the other aims to destroy trust in government institutions. Without the separate dimensions for story, group positioning, and goal, these critical differences would be invisible. The researchers also built a formal computer check to ensure the logic of their system held together, confirming that the rules for deriving higher-level labels from basic evidence were consistent and unbreakable.
This new framework is not a tool for deciding if a message is true or false, nor is it a way to prove who created it or whether it actually changed anyone's mind. It is a measurement tool, a way to describe the architecture of influence so that it can be studied, compared, and understood. The researchers acknowledge that applying this system to the millions of messages generated every day will require computers to do the heavy lifting, with humans guiding the process and checking the results. They also note that the system was developed primarily using English-language political examples, and that adapting it to other cultures will require careful work to understand local cues and norms. However, the core insight remains: to understand how we are being influenced in an age of artificial intelligence, we must look past the question of truth and examine the structure of the message itself. By mapping the form of influence, we gain the ability to see the machinery behind the curtain, distinguishing between a simple statement of fact and a carefully constructed tool designed to move us.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.